Finding of Agr Phase Variants in Staphylococcus aureus

Staphylococcus aureus is responsible for a broad range of infections. This pathogen has a vast arsenal of virulence factors at its disposal, but avirulent strains are frequently isolated as the cause of clinical infections. These isolates have a mutated agr locus and have been believed to have no evolutionary future. Here we show that a fraction of Agr-negative strains can repair their mutated agr locus with mechanisms resembling phase variation. The agr revertants sustain an Agr OFF state as long as they exist as a minority but can activate their Agr system upon phagocytosis. These revertant cells might function as a cryptic insurance strategy to survive immune-mediated host stress that arises during infection.


RESULTS
Serial passaging of S. aureus generates Agr-negative variants. By their nature, phase variants emerge spontaneously over time with successive generations of growth. To isolate in vitro Agr-negative S. aureus, we serially passaged strains MW2 and s0437 and screened for nonhemolytic colonies on sheep blood agar (SBA). Hemolysis on SBA was used as the first screen, as it provided a quick method to distinguish potential Agr-negative colonies from a large number of samples. MW2 and s0437 were passaged in tryptic soy broth (TSB), TSB containing 5% human blood, or TSB containing 5% human serum. Nonhemolytic colonies did not appear in some experiments even after several passages, but in others a large number were detected. Namely, the emergence of nonhemolytic colonies fluctuated greatly due to their spontaneous nature. All growth conditions allowed for generation of nonhemolytic variants. In total, 209 out of 14,948 MW2 colonies and 4 out of the 308 s0437 colonies were nonhemolytic.
Agr-negative variants can phenotypically and genotypically revert to the wild type. We then tested if the isolated Agr-negative variants could revert their Agr activity over multiple generations. We carried out successive passages of 30 isolated Agrnegative variants and plated serial dilutions of the cultures onto SBA to check for the generation of any hemolytic colonies. Two agr-negative variants {termed RAV MW2 (reversible Agr variant [AV] derived from MW2) and RAV s0437 (from s0437)} were able to reproducibly give rise to hemolytic colonies. To verify that these colonies were Agr revertants, we carried out CAMP tests and confirmed the presence of the characteristic ␦ hemolysin arrowhead in the revertant colonies that was absent in the Agr-negative variants (Fig. 1). The average frequencies of revertants after 4 successive cultures (i.e., the number of hemolytic colonies compared to nonhemolytic colonies on SBA plates) were ϳ2% in RAV MW2 and ϳ0.8% in RAV s0437 (Table 1). We also noted that RAV MW2 generated 4-fold-higher numbers of rifampin-resistant colonies (median ϭ 2.3 ϫ 10 Ϫ8 and mean ϭ 2.7 ϫ 10 Ϫ8 [n ϭ 10]) after one overnight culture than did MW2 (median ϭ 5.4 ϫ 10 Ϫ9 and mean ϭ 6.8 ϫ 10 Ϫ9 [n ϭ 8]) (P Ͻ 0.01, in a two-tailed t test), although this study did not further address if the higher mutation rate is responsible for the reversion rate of RAV MW2 . The background mutation rates (measured by rifampin resistance) in the s0437 wild-type (WT) strain (median ϭ 1.4 ϫ 10 Ϫ7 and mean ϭ 1.4 ϫ 10 Ϫ7 [n ϭ 10]) and RAV s0437 (median ϭ 1.0 ϫ 10 Ϫ7 and mean ϭ 9.7 ϫ 10 Ϫ8 [n ϭ 10]) were also determined.
Sequencing of the agr loci of the variants (both the reversible isolates and some of the irreversible isolates) and revertants uncovered underlying genetic mechanisms for Agr shutdown (Fig. 2). RAV MW2 had a duplication and inversion in agrC. The agr locus in the RAV MW2 revertant was identical to that of the WT. RAV s0437 had an alteration in a poly(A) tract within agrA, which reverted to the WT sequence in the RAV s0437 revertant. Most of the sequenced irreversible Agr-negative MW2 isolates had insertions and/or point mutations across the agr locus, in line with previous studies (8)(9)(10). Although one irreversible isolate (AV3 MW2 ) had a duplication within agrA, we were unable to generate any revertants. Taken together, these results show that a fraction of the Agr-negative strains isolated in vitro can phenotypically revert their Agr activity and that there are multiple underlying mutational events behind this reversion.
We also carried out the revertant generation procedure on 61 clinically isolated Agr-negative strains. Although some strains showed a phenotypic Agr reversion in some experiments, only one strain, an MRSA isolate designated RAV(66r), was reproducibly reversible (Table 1). However, the detection limit of the initial screening was ϳ3.0 ϫ 10 Ϫ3 , and thus, the possibility remains that more reproducibly reversible strains could be identified using a higher detection limit. The mean frequency of reversion in RAV(66r) was 0.12% (Table 1). Upon sequencing of the agrA and -C loci of RAV(66r) and the RAV(66r) revertant, a reversible poly(A) tract alteration was found in agrA (Fig. 2). The average background mutation rate for RAV(66r) was 7.0 ϫ 10 Ϫ9 (median ϭ 5.4 ϫ 10 Ϫ9 [n ϭ 10]).
A summary of all the Agr variants and revertants generated or collected in this study, along with the growth conditions used for their generation as well as the origin strain, can be found in Table 1.
A minor population of agr-intact cells does not activate its own Agr system in a non-autoinducing peptide (AIP)-producing population but can through local segregation on solid media. Due to its quorum sensing nature and the low reversion FIG 1 Serial passaging of Agr-negative strains can give rise to colonies with reverted Agr activity. Shown are flowcharts describing the screening procedure used to isolate strains in this study. (A) Parental strains MW2 and s0437 were passaged in successive liquid cultures, followed by screening for nonhemolytic colonies. CAMP tests were carried out on the nonhemolytic colonies alongside the parent strains to confirm loss of Agr activity. The CAMP image depicts a nonhemolytic isolate generated from MW2. (B) Agr-negative variants were passaged in successive liquid cultures, followed by screening for hemolytic colonies. CAMP tests were carried out on hemolytic colonies alongside the reversible Agr-variant (RAV) strains and the original parental strains. The CAMP images depict RAV MW2 and RAV s0437 alongside their respective revertant and WT stains. The Agr-negative MRSA clinical isolate RAV(66r) and the RAV(66r) revertant are also shown. MW2 Δagr was used as a negative control. To test this, Agr activity was monitored using a venus reporter system under the control of the agr P3 promoter in populations of an isogenic Δagr strain unable to produce any AIP (MW2 ⌬agr P3venus) with various percentages of agr-intact cells (MW2 P3venus) mixed in, as well as in populations of the reversible RAV MW2 . Venus fluorescent protein is stable enough to detect even transient activation of a promoter at the endpoint of the experiment. When spotted onto TSA with 12.5 g/ml of chloramphenicol (Cm12.5), no fluorescence was observed in the spotted areas of samples with an MW2 P3venus content less than 10% (Fig. 3A). However, we were able to visualize streaks of localized fluorescence on the outer edges of spotted areas in samples with an MW2 P3venus content as low as 0.1% (Fig. 3A). This suggests that a minor population of agr-intact cells can activate their Agr system upon propagation on solid media, likely because the AIP concentration in the locality can exceed the threshold. A similar pattern of fluorescence could be detected in RAV MW2 but not in the irreversible strain AV1 MW2 (Fig. 3B). RAV MW2 , which carries a mutation in agrC, would still be able to Agr Phase Variants ® produce basal levels of AIP, but the local signals in the RAV MW2 colony suggested the presence of a minor population of agr-intact revertant that were generated through colony expansion. Figure 3C shows a control reporter system expressing gfp under the control of the SigB-dependent asp promoter in N315 and its isogenic ΔsigB strain. Green fluorescent protein (GFP)-positive cells remained detectable within the entire spotted area of samples with low N315 content (Fig. 3C), confirming that the quorum-independent reporter system can be detected in the same experimental settings.
When the proportion of Agr-positive cells in liquid culture corresponded to the observed reversion frequency, i.e., Ͻ10%, reporter activity was undetectable during the   (Fig. 4A). In contrast, when the proportion of WT cells was increased to 10%, reporter expression was observed, suggesting that this is the threshold that is necessary for Agr activation under planktonic conditions. The frequency of SigB active cells remained constant and detectable for all samples over the time course tested (Fig. 4B). To eliminate the possibility that the agr-intact cells had been lost during the time course tested, we plated the samples on SBA after the experiment and confirmed that the frequency of hemolytic cells had not changed from the start of the experiment (Fig. 4C). Agr Phase Variants ® Taken together, these data indicate that a minor population of Agr revertant cells does not activate its Agr system when growing in planktonic conditions. Phagocytosis triggers expression of the Agr system in revertant cells. We showed that agr-intact cells sustain the Agr OFF state when they are a minority (Ͻ10%) in an Agr-negative population. It is known that a single agr-intact cell can activate its own Agr system when it is packed into a narrow matrix mimicking the phagosome environment (25) in a phenomenon termed "diffusion sensing" (26,27) and that Agr activation is induced upon phagocytosis (28,29). Therefore, the phagosome environment must allow agr revertant cells to switch the Agr system ON.
To test this, a phagocytosis assay on RAV MW2 P3venus was carried out. An RAV MW2 overnight culture contained minor revertant cells (Fig. 5C, RAV MW2 , Before infection). These minor revertant cells were negative in GFP expression in liquid culture without macrophages (Fig. 5A, ϪM). As expected, fluorescent cells were detected within macrophages at 4 and 8 h postinfection (Fig. 5A and B). To test whether the presence of macrophages affected the agr reversion frequency, serial dilutions of

DISCUSSION
In this study, we showed that a part of Agr-negative isolates generated in vitro, as well as those isolated from patients, can phenotypically revert their Agr activity with underlying genetic mechanisms reminiscent of phase variation. In addition, we showed that while revertant cells are generated at a low frequency, allowing them to sustain an Agr OFF state in planktonic culture, they can activate their Agr system upon phagocytosis. This suggests a cryptic strategy for the bacterial population to survive phagocytic stress and allow for continued infection (Fig. 6).
Agr-negative strains are commonly isolated from patients and are linked to more severe infection outcomes, being associated with higher mortality and increased duration of bacteremia (11,14). In fact, Agr-negative strains have been shown to have a higher propensity in causing bacteremia, and this has been suggested as being due to a lower energy burden (13) conferring a fitness increase (15). In general, infections arising from Agr-negative S. aureus are considered a "dead end," as interhost spread of the bacteria is impaired and thus the infecting strain lineage goes extinct (12,15). According to the "evolutionary trade-offs" concept proposed by Laabei et al., highly toxic S. aureus isolates that tend to be associated with superficial skin and soft tissue infections have high interhost fitness and are maintained at the epidemiological level, while low-toxicity isolates that have propensity to cause invasive infections (bacteremia) are impaired in interhost spread (15). On the other hand, recent studies carried out in mouse models suggest that Agr-negative S. aureus can spread between hosts via the gastrointestinal tract (30), and thus, Agr dysfunction may result in complexity, rather than outright abolishment, of passage to the next host. It is interesting that non-toxinproducing strains have also been linked to the shift from commensalism to infection Agr Phase Variants (15,31). In this context, the phase variation-mediated shutdown of the Agr system adds an extra layer of complexity for S. aureus.
Neutrophils and macrophages are key elements in the host innate immune response to S. aureus infection. It has long been known that S. aureus can survive within phagosomes of these immune cells in an Agr-dependent manner (28,29) without compromising the viability of the phagocytes through upregulation of antiapoptotic factors (32,33). There are suggestions that phagocytic cells may act as a "Trojan horse" by facilitating dissemination of S. aureus to different infectious sites (34-36). Gresham et al. showed that S. aureus-containing polymorphonuclear neutrophils can successfully establish infection in naive mice (37), and Lehar et al. demonstrated that intracellular MRSA, sequestered within macrophages and neutrophils, was more successful at colonizing organs such as the kidney and brain (as well as being protected against vancomycin) than planktonic bacteria (33). The Agr-reverted phase variants identified in this study may play a role in this immune-mediated dissemination; Agr revertants would be able to survive phagocytosis and would hide within the Trojan horse as the phagocytic cells migrate away from the site of infection.
In summary, we show here that a fraction of Agr-negative strains are phase variants that arise by reversible genetic mutations in the agr locus either by alterations within short sequence repeats (SSR) or by site-specific rearrangements. To our knowledge, this is the first report demonstrating that Agr-negative strains can revert their Agr phenotype. Further, we discuss the possibility that the revertant cells serve as a cryptic strategy against the dead-end nature of infections favored by Agr-negative strains. These findings take us a step forward in understanding the infectious lifestyle of S. aureus and provide a new outlook on the prevalence of Agr-negative strains in the clinical setting. More extensive studies with clinically relevant samples, along with in-depth genotyping of any Agr-negative clinical samples identified as being reversible, are required to assess the prevalence and impact of Agr reversion in the health care setting.

MATERIALS AND METHODS
Bacterial strains and growth conditions. Bacterial strains and plasmids used in this study are listed in Table 2. Strains were routinely grown in tryptic soy broth (TSB) at 37°C with shaking at 180 rpm. A total of 12.5 g/ml of chloramphenicol (Cm12.5) was added when necessary.
Strains MW2 and s0437 were used to generate in vitro Agr-negative mutants and to test their reversibility. Strain N315 was used in some fluorescence experiments. A total of 173 clinical S. aureus samples, comprising 74 MRSA and 99 methicillin-susceptible S. aureus (MSSA) isolates, were collected from the Kanto region of Japan. Cells were picked from pure cultures on Mueller-Hinton agar plates and grown in TSB before making glycerol stocks.
Construction of agr deletion mutant. An agr deletion mutant was constructed by double-crossover homologous recombination using the pMAD-tet vector as described previously (38,39). Briefly, the upstream and downstream regions flanking the entire agr locus were amplified using primer pairs agr-up-f(BamHI) and agr-up-r(BamHI) and agr-down-f(ecoR1) and agr-down-r(bgl2) ( Table 3) and then cloned into pMAD-tet to generate pMAD-tet-Δagr. This plasmid was introduced into strain MW2 by phage transduction after passaging through strain RN4220, and the agr mutant was constructed as previously described (39). The absence of the agr locus was confirmed by lack of hemolysis.
Variant generation. Agr-negative variants were generated from two S. aureus strains: MW2 and s0437. Independent cultures of MW2 and s0437 were grown overnight in TSB to stationary phase and then subcultured at a 1,000-fold dilution into fresh TSB, TSB with 5% human blood, or TSB with 5% serum. Blood was mixed in a 3:2 ratio with filter-sterilized anticoagulation solution. The anticoagulation solution consisted of 0.32% (wt/vol) citric acid monohydrate, 0.88% (wt/vol) trisodium citrate dihydrate, and 0.88% (wt/vol) glucose prepared in Millipore water. Subcultures were repeated up to 3 to ϳ7 times, and samples were plated onto sheep blood agar (SBA) at each stage to screen for nonhemolytic colonies.
Revertant generation. Agr-negative variant strains derived from MW2 and s0437 were grown overnight in TSB in 3 independent cultures to stationary phase. Samples were subcultured at a 1,000-fold dilution into fresh TSB and grown to stationary phase. Subcultures were repeated up to 4 times, and serial dilutions of the samples were plated onto SBA at each point to screen for hemolytic colonies.
A total of 61 of the 173 clinical isolates were characterized as being Agr negative by the CAMP test. All 61 Agr-negative strains were tested in an initial revertant generation screening. A single culture of each strain was grown overnight in TSB to stationary phase. Samples were subcultured at a 1,000-fold dilution into fresh TSB and grown to stationary phase. Subcultures were repeated up to 4 times, and serial dilutions of the samples were plated onto SBA at each point to screen for hemolytic colonies.
Assessment of Agr activity on sheep blood agar plates. Traber et al. demonstrated that a variation of the CAMP test can be used to assess Agr activity of S. aureus by streaking test samples perpendicularly to an S. aureus strain that produces only ␤ hemolysin and analyzing the resulting patterns of hemolysis for the distinct pattern of Agr-controlled ␦ hemolysin (8,40). Thus, the CAMP test was used to assess Agr activities of the isolated variant and revertant strains. An overnight culture of strain RN4220 (produces only ␤ hemolysin) grown in TSB was streaked down the center of an SBA plate using a cotton swab, and the plate was incubated at 37°C for 4 to 6 h. Following this, test samples and controls (strain MW2 and its isogenic Δagr mutant) were streaked perpendicularly to RN4200, approaching the streak but not coming into direct contact with it. Plates were incubated at 37°C for 12 to 16 h, until the hemolysis patterns of the controls became clearly visible. Further incubation at 4°C was carried out, if necessary, to enhance hemolysis. agr locus sequencing. Genomic DNA was purified using standard procedures. The whole agr locus encompassing hld (NCBI gene identifier [ID] 1004072, Staphylococcus aureus subsp. aureus MW2) through agrA (NCBI gene ID 1004076, Staphylococcus aureus subsp. aureus MW2) was amplified by PCR using the primers agr front and agr back (Table 3) and subjected to direct sequencing using the agr4, agr5, and agr back primers (Table 3) (Fasmac, Japan). The sequence data were analyzed using the DNASTAR sequence analysis suite.
Plasmid construction. The pRITP3 (RNAIII) -venus Agr reporter plasmid was constructed by Life Technologies, Japan. This plasmid has the P3 promoter region of RNAIII (Ϫ181 to ϩ12 from the transcription initiation site), the venus fluorescent protein coding region, and the terminator region from clpB cloned into the pRIT5H backbone (41).
Measurement of reversion frequency. Glycerol stocks of MW2, RVA MW2 , MW2 Δagr, and RAV(66r) were inoculated in fresh TSB in 10 separate test tubes. Test tube cultures were grown to stationary phase and then subcultured at a 1,000-fold dilution into fresh TSB. Subcultures were successively repeated 4 times in total. Serial dilutions from the final stationary-phase subculture were plated on SBA, and plates were incubated for 24 h at 37°C, after which the reversion frequency was scored as the percentage of hemolytic colonies on the final SBA plates.
RAW 264.7 cells were seeded in a 16-well plate (0.5 ϫ 10 4 cells/well, 200 l) and were grown overnight until reaching a density of 1 ϫ 10 4 per well. The cell culture medium was replaced with drug-free RPMI 1640 medium supplemented with glutamine and 5% FBS 1 h before the infection.
MW2 P3venus, RAV MW2 P3venus, and MW2 Δagr P3venus were grown overnight in TSB with Cm12.5. Cells were harvested by centrifugation (6,000 rpm, 2 min, and 4°C) from overnight culture and resuspended in RPMI 1640 medium. Cell density was adjusted to 1 ϫ 10 9 CFU/ml by referencing the OD 600 value. RAW 264.7 cells were infected with 10 l of the bacterial cells at a multiplicity of infection (MOI) of ϳ10. At 1 h postinfection, the culture medium was replaced with RPMI 1640 medium supplemented with glutamine, 5% FBS, and gentamicin (50 g/ml) and the samples were incubated for 1 h to kill nonphagocytosed bacteria. Following this, the medium was once again replaced with drug-free RPMI 1640 medium supplemented with glutamine and 5% FBS. Samples were prepared for microscopy at 1, 5, and 9 h postinfection by removing the supernatant and fixing the cells with 200 l of methanol. Images were captured using a light microscope (FSX100; Olympus, Center Valley, PA) and viewed with FSX-BSW software (Olympus). To check whether the presence of RAW 264.7 cells affected Agr reversion frequency, bacteria were collected from a separate set of samples after the final time point by lysing macrophages with 200 l of cold PBS supplemented with 0.02% Triton X-100 and mixing with a pipette. These samples were then plated on SBA and the frequency of reversion was assessed by the frequency of hemolytic colonies and compared to that of the preinfection samples.